1 Model-Driven Architecture: Vision, Standards And Emerging Technologies Position Paper Submitted to ECOOP 2001 Workshop on Metamodeling and Adaptive Object Models John D. Poole Hyperion Solutions Corporation April 2001 [email protected] Poole -- Model-Driven Architecture. ECOOP 2001. 2 1. Introduction Recently, the Object Management Group introduced the Model-Driven Architecture (MDA) initiative as an approach to system-specification and interoperability based on the use of formal models [MDA, MDA2, DSouza]. In MDA, platform-independent models (PIMs) are initially expressed in a platform-independent modeling language, such as UML. The platform-independent model is subsequently translated to a platform-specific model (PSM) by mapping the PIM to some implementation language or platform (e.g., Java) using formal rules. At the core of the MDA concept are a number of important OMG standards: The Unified Modeling Language (UML), Meta Object Facility (MOF), XML Metadata Interchange (XMI), and the Common Warehouse Metamodel (CWM). These standards define the core infrastructure of the MDA, and have greatly contributed to the current state-of-the- art of systems modeling [MDA2]. As an OMG process, the MDA represents a major evolutionary step in the way the OMG defines interoperability standards. For a very long time, interoperability had been based largely on CORBA standards and services. Heterogeneous software systems inter- operate at the level of standard component interfaces. The MDA process, on the other hand, places formal system models at the core of the interoperability problem. What is most significant about this approach is the independence of the system specification from the implementation technology or platform. The system definition exists independently of any implementation model and has formal mappings to many possible platform infrastructures (e.g., Java, XML, SOAP). The MDA has significant implications for the disciplines of Metamodeling and Adaptive Object Models (AOMs). Metamodeling is the primary activity in the specification, or modeling, of metadata. Interoperability in heterogeneous environments is ultimately achieved via shared metadata and the overall strategy for sharing and understanding metadata consists of the automated development, publishing, management, and interpretation of models.1 AOM technology provides dynamic system behavior based on run-time interpretation of such models. Architectures based on AOMs are highly interoperable, easily extended at run-time, and completely dynamic in terms of their overall behavioral specifications (i.e., their range of behavior is not bound by hard-coded logic). The core standards of the MDA (UML, MOF, XMI, CWM) form the basis for building coherent schemes for authoring, publishing, and managing models within a model-driven architecture. There is also a highly complementary trend currently building within the industry toward the realization of these MDA standards in the Java platform (i.e., standard mappings of platform-independent models to platform-dependent models, where the platform-dependent model is the Java platform). This is a sensible implementation strategy, since development and integration is greatly facilitated through common 1 In this context, the terms model and metadata can be used interchangeably, although model would seem to have a more general connotation. Poole -- Model-Driven Architecture. ECOOP 2001. 3 platform services and programming models (interfaces or APIs), provided as part of the Java platform. Java 2 Platform, Enterprise Edition (J2EE), has become a leading industry standard for implementing and deploying component-based, distributed applications in multi-tier, Web-centric environments. Current efforts within the Java Community Process to develop pure Java programming models realizing OMG standards in the form of J2EE standard APIs (i.e., JMI, JOLAP and JDM) further enhance the metadata-based interoperability of distributed applications. This paper surveys the core OMG MDA standards (i.e., UML, MOF, XMI and CWM) and discusses the current attempts at mapping these standards to J2EE, as examples of PIM-to-PSM translations that are currently under development. These forthcoming APIs will provide the initial building blocks for a new generation of systems based on the model-driven architecture concept. The progression of these initial MDA realizations to AOMs is the next logical step in this evolution. 2. The Vision This paper's proposed vision for the evolution of the MDA is twofold, consisting of both a near term vision and a long term vision of the future. The near term vision (i.e., nearly seamless interoperability, based on formal PIM-PSM translations and shared metadata) is achievable right now. The supporting technologies are largely specified and implementations are currently being built by a number of organizations. The long-term vision (i.e., the wide spread deployment of AOMs, as an evolution of the MDA), on the other hand, is still being conceptualized. Overview of the Near Term Vision The proposed near term vision is that of an environment in which efficient and nearly seamless interoperability between diverse applications, tools and databases is achieved through the interchange of shared models. Components participating in this environment leverage standard services provided by implementations of MDA standards that enable them to expose and interchange their metadata as instances of well-defined models. These platform services have standard definitions that are expressed via standard programming models (APIs), which are automatically generated from platform- independent models. The Importance of Shared Metadata Metadata is critical to all aspects of interoperability within any heterogeneous environment. In fact, metadata is the primary means by which interoperability is achieved (interoperability is largely facilitated by standard APIs, but ultimately requires shared metadata as the definitions of system semantics and capabilities). Any MDA- based system must have the ability to store, manage and publish both application- and system-level metadata (including descriptions of the environment itself). Applications, tools, databases, and other components plug into the environment and discover metadata Poole -- Model-Driven Architecture. ECOOP 2001. 4 descriptions pertaining to the environment. Similarly, a component or product introduced into the environment can also publish its own metadata to the rest of the environment. This scenario is illustrated in Fig. 1. Metadata Modeling & Authoring Visual Modeler Client Application Analysis Tool RDBMS Metadata Service Metadata & Data Metadata & Data Metadata & Data Metadata Common Services / Programming Models (APIs) Model-Driven Architecture Platform (Infrastructure) Figure 1: An Example of a Realization of Model-Driven Architecture Having an abundance of shared, descriptive metadata (ubiquitous metadata) facilitates software interoperability between platform components in very specific ways, including: · Data interchange, transformation, and type mapping between dissimilar data resources can be driven by formal, product-independent metadata descriptions of the transformations, data types, and type-system mappings. · Schema generation can be based on shared metadata descriptions of common schema elements. For example, both a relational database and an OLAP server can build their own internal representations of a dimensional model, according to a standard, metadata-based definition of "dimension" published to the environment. Having common metadata definitions facilitates data interchange between subsystems, because there is a common understanding of what the data means. · Business intelligence and visualization functions can utilize metadata in the processing and formatting of data for analysis and display. Metadata descriptions confer the "higher level of meaning" on data items that analysts and reporting users need in order to make sense of data points and results (e.g., definitions of business terms, glossaries, taxonomies, nomenclatures, etc., are a part of the shared metadata). · Software components with no prior knowledge of each other's capabilities, interfaces, and data representations can interoperate once they've established a metadata "handshake", in which each exposes its features and assesses those of the other. Note that this exchange of knowledge does not always need to be complete, but to the extent that components can make sense of each other's capabilities, they can interact with one another. In the absence of specific knowledge, components might rely on Poole -- Model-Driven Architecture. ECOOP 2001. 5 standard defaults, or may be able to refer to some other source of information to fill the knowledge gaps. An MDA-based system does not require that internal representations of metadata within applications, tools, and databases be modified to correspond to the shared definitions. Product-specific internals and programming models remain as they are. Shared metadata consists of externalized definitions that are interchanged between participating components. These external definitions are readily understood by components that agree on the metamodel describing the metadata (e.g., CWM). External definitions are highly generic, but also possess sufficient semantic completeness (with respect to the problem domains that components need to address), and are, therefore, understood by a wide range of participants. Highly product-specific metadata that does not fit